AMD's Lisa Su: 'Economic moat' does not apply to AI industry

Amanda Liang, Taipei; Jack Wu, DIGITIMES Asia 0

AMD CEO Lisa Su. Credit: AFP

Previously, at the Code Conference held in California, AMD CEO Lisa Su discussed topics covering the AI industry, expansion of domestic semiconductor production in the US, Huawei's new smartphone, and the market competition with Nvidia. During the event, when asked about the "economic moat" in the AI sector, Su expressed that she's "not a believer in moats" in a market that's moving as fast as AI.

The term "economic moat" refers to "a business's ability to maintain competitive advantages." Popularized by Warren Buffett, it has since become an important concept when evaluating the operational performances of corporations. For instance, TSMC has a global market share of 90% in advanced processes, giving it an absolute advantage over competitors like Samsung Foundry and IFS.

Nvidia's leading status as the leader in the AI sector has had some people asking if it's creating an economic moat in the sector and whether or not AMD can overcome it. However, Su believes that no company has been able to form a moat in the current AI market due to how fast the sector is developing. The scale of the market is also growing at wasp speed. Andrew Ng, who visited Taiwan recently, predicted that the generative AI scale is going to double in three years. In other words, no AI chip company currently possesses an absolute competitive advantage over its peers.

Su believes that competition in the AI industry is more open-ended than what outsiders perceive. Development in the next decade will be significantly different from the past decade, and the development method of AI is also changing. "The benefit of an open approach is that there's no one company that has all of the ideas," she stated.

Su admitted that Nvidia's H100 has a competitive edge in terms of AI training workloads. However, the AI market is not a binary one where one chip product is clearly better. Some products excel in AI training, while other products may be more suitable for AI inference. This entirely depends on how manufacturers combine these components.

Regarding AMD's progress with MI300, Su emphasized that they have created an outstanding AI inference product, particularly for LLM inference tasks. Looking ahead AMD believes that AI inference will create an even larger market demand. The AI market is changing rapidly and generative AI is still in its early market introductory stage. Because of this, Su stated that she doesn't think that the AI market already has a major economic moat.

When asked about the rumored 7nm chip in Huawei's new product, Su responded that she doesn't consider this a major geopolitical event. From the US's perspective, the government has spent a considerable amount of time hoping to secure its technological edge over China. With how important technology is to national security, this stance is understandable.

However, she viewed the discussion around Huawei's new chip from the perspective of AMD, a global corporation. For AMD, China is a crucial market. Compared to other products, AMD indeed sold more consumer-related products to the Chinese market. Therefore, AMD has the opportunity to address some geopolitical issues with a balanced approach.